Lawrence Berkeley National Laboratory Recent Work

Title Solar splitting by photovoltaic-electrolysis with a solar-to- efficiency over 30%

Permalink https://escholarship.org/uc/item/5wr3t7qs

Journal Nature Communications, 7

Authors Jia, J Seitz, LC Benck, JD et al.

Publication Date 2016-10-31

DOI 10.1038/ncomms13237

Peer reviewed

eScholarship.org Powered by the California Digital Library University of California ARTICLE

Received 6 May 2016 | Accepted 9 Sep 2016 | Published 31 Oct 2016 DOI: 10.1038/ncomms13237 OPEN Solar water splitting by photovoltaic-electrolysis with a solar-to-hydrogen efficiency over 30%

Jieyang Jia1,*, Linsey C. Seitz2,*, Jesse D. Benck2,*, Yijie Huo1, Yusi Chen1, Jia Wei Desmond Ng2,3, Taner Bilir4, James S. Harris1 & Thomas F. Jaramillo2

Hydrogen production via electrochemical water splitting is a promising approach for storing solar energy. For this technology to be economically competitive, it is critical to develop water splitting systems with high solar-to-hydrogen (STH) efficiencies. Here we report a photo- voltaic-electrolysis system with the highest STH efficiency for any water splitting technology to date, to the best of our knowledge. Our system consists of two polymer electrolyte membrane electrolysers in series with one InGaP/GaAs/GaInNAsSb triple-junction , which produces a large-enough voltage to drive both electrolysers with no additional energy input. The solar concentration is adjusted such that the maximum power point of the photovoltaic is well matched to the operating capacity of the electrolysers to optimize the system efficiency. The system achieves a 48-h average STH efficiency of 30%. These results demonstrate the potential of photovoltaic-electrolysis systems for cost-effective solar .

1 Department of Electrical Engineering, Stanford University, 350 Serra Mall, Stanford, California 94305, USA. 2 Department of Chemical Engineering, Stanford University, 443 Via Ortega, Shriram Center Room 305, Stanford, California 94305, USA. 3 Institute of Chemical and Engineering Sciences, Agency for Science, Technology and Research, Jurong Island 627833, Singapore. 4 Solar Junction, 401 Charcot Avenue, San Jose, California 95131, USA. * These authors contributed equally to this work. Correspondence and requests for materials should be addressed to J.S.H. (email: [email protected]) or to T.F.J. (email: [email protected]).

NATURE COMMUNICATIONS | 7:13237 | DOI: 10.1038/ncomms13237 | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13237

he sustainable nature of solar electricity along with its potential for building extremely high-efficiency solar H2 produc- associated large resource potential and falling costs have tion systems using current state-of-the-art commercially available Tmotivated a rapid increase in the deployment of utility- solar cells and laboratory PEM electrolysers. The device design scale solar electricity generation plants in recent years1. As the presented herein could provide a viable route to implementing installed capacity of (PVs) continues to grow, large-scale solar water splitting installations. cost-effective technologies for solar energy storage will be critical to mitigate the intermittency of the solar resource and to Results maintain stability of the electrical grid2. Hydrogen generation via PV-electrolysis system design. A schematic of the PV-electrolysis solar water splitting represents a promising solution to these system is shown in Fig. 1. The solar cell is a commercially challenges, as H2 can be stored, transported and consumed available triple-junction solar cell manufactured by Solar Junc- without generating harmful byproducts3–8. However, the cost of tion, with an active area of 0.316 cm2. From top to bottom, the H2 produced by electrolysis is still significantly higher than that three subcells of the PV are made of InGaP (Eg ¼ 1.895 eV), GaAs 31 produced by fossil fuels. The Department of Energy has (Eg ¼ 1.414 eV) and GaInNAs(Sb) (Eg ¼ 0.965 eV) respectively . calculated the H2 threshold cost to be $2.00–$4.00 per gallon of The cell was installed on a water-cooled stage, which maintained 9 gasoline equivalent , whereas the most up-to-date reported H2 a cell temperature of B25 °C and was illuminated with white production cost via electrolysis is $3.26–$6.62 per gallon of light from a xenon arc lamp to simulate concentrated AM 1.5D gasoline equivalent10. There are several promising approaches to solar illumination. (See Supplementary Fig. 1 for comparative large-scale solar water splitting, including photochemical, xenon arc lamp and AM 1.5D spectra.) The two PEM electro- photoelectrochemical (PEC) and PV-electrolysis systems8; none lysers consist of Nafion membranes coated with 0.5 mg cm 2 Pt of these approaches are currently economically viable compared black catalyst at the cathode and 2 mg cm 2 Ir black catalyst at with today’s technologies3,6,11. the anode. The two electrolysers were connected in series with the To be practical for large-scale deployment, the cost of solar H2 PV cell and a potentiostat, which was used to measure the current generation must be significantly reduced. Previous studies through the circuit. Water was pumped into the anode have predicted that achieving a high solar-to-hydrogen (STH) compartment of the first electrolyser, whereas the cathode of efficiency is a significant driving force for reducing the H2 the first electrolyser had no input flow. The water and O2 effluent generation cost12–14. To date, the highest efficiency demonstrated from the first electrolyser’s anode compartment flowed into the using a PEC water splitting system with at least one anode compartment of the second electrolyser. Likewise, the H2 semiconductor–liquid junction is 12.4% (refs 15,16). Theoretical from the cathode side of the first electrolyser flowed into the studies using a variety of assumptions have predicted that the cathode side of the second electrolyser. The H2 and O2 products maximum attainable efficiency using a tandem PEC water flowing out of the second electrolyser were collected and splitting device is 23–32% (refs 17–20). PV-electrolysis systems quantified, whereas the unreacted water was fed back into a have demonstrated exceptional potential to achieve even higher water reservoir and recycled through the system. The temperature STH efficiencies8,11,21–26. The highest STH efficiency of the electrolysers was held at B80 °C, consistent with standard demonstrated to date, 24.4%, was delivered by a PV-electrolysis operating conditions for industrial water electrolysers21,22. The system using GaInP/GaAs/Ge multi-junction solar cells and system operated continuously for 48 h without interruption. polymer electrolyte electrochemical cells24. For comparison, the Further details of the experimental setup are included in the best multi-junction PV created to date demonstrated a solar-to- Methods section. electricity conversion efficiency of 46.0% under concentrated 27 illumination . In theory, PV-electrolysis systems could poten- System performance. The I–V characteristics of the cell under tially achieve up to 90–95% of the PV efficiency, which could B both 1 sun and concentrated illumination are shown in Fig. 2. allow for PV-electrolysis efficiencies of 57% for a 3J cell and Following a standard method for characterizing concentrated PV B62% for a 4J or 5J cell28. These values indicate that there is significant room for further improvement in the performance of PV-electrolysis system prototypes. Electrolysers H2O The discrepancy between reported STH efficiencies for PV- Unreacted H2O H & O H O recirculated electrolysis devices and stand-alone solar-to-electricity PV 2 2 2 products O2 efficiencies mainly arises from poor matching of the current– H2 – – quantified e e e– voltage (I–V) characteristics of multi-junction PVs with those of 8 water electrolysers . The maximum power-point voltage (VMPP) of a typical commercial triple-junction solar cell is in the range of 2.0–3.5 V under 1–1,000 suns of illumination. However, the h+ h+ thermodynamic minimum voltage required to electrolyse water is only 1.23 V at 300 K (refs 7,29), with practical operating voltages e–

in the range of 1.5–1.9 V (refs 7,30). Electrolysing water using a Curent collector Liquid/gas flow Ir black Nafion / Nafion membranePt black NafionCarbon / paper H2O Ti mesh + O – voltage in excess of the thermodynamic minimum voltage results 2 e in energy wasted as heat rather than stored in H2 chemical bonds. H Previously, this limitation was overcome by coupling multiple PV 2 and/or electrolyser units in series, to optimize the match between 23,24 InGaP (1.9 eV) the voltage characteristics of these device components , GaAs (1.4 eV) although the efficiencies achieved were still far from optimal. GaInNAs(Sb) (1.0 eV) In this work, we employ a high-efficiency triple-junction solar cell with two series-connected polymer electrolyte membrane e– Photovoltaic (PEM) electrolysers to achieve very high STH efficiency. Our system produces H2 with a 48 h average STH efficiency of 30%, Figure 1 | PV-electrolysis device schematic. The PV-electrolysis system the highest efficiency reported to date for any solar H2 production consists of a triple-junction solar cell and two PEM electrolysers connected system, to the best of our knowledge. This work demonstrates the in series.

2 NATURE COMMUNICATIONS | 7:13237 | DOI: 10.1038/ncomms13237 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13237 ARTICLE

a 5.0 b 200 4.5 600 14 180 ) ) –2 4.0 –2 12 160 500 3.5 140 10 3.0 1 Sun 400

42 Suns (mA cm 2 120 2 Area = 0.316 cm y 2.5 8 Area = 0.316 cm V = 2.79 V 100 OC VOC = 3.21 V 300 2.0 J = 13.9 mA cm–2 –2 SC 6 80 JSC = 584 mA cm Current (mA) Current (mA) VMPP = 2.48 V V = 2.91 V 1.5 –2 60 MPP 200 J = 13.4 mA cm –2 MPP 4 JMPP = 566 mA cm 1.0 P = 33.2 mW cm–2 40 –2 Current density (mA cm MPP PMPP = 1,640 mW cm Current densit 2 100 0.5 Fill Factor = 0.858 20 Fill Factor = 0.872 0.0 0 0 0 0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 Voltage (V) Voltage (V)

Figure 2 | PV cell performance. The I–V characteristics of the triple-junction solar cell under (a) 1 sun and (b) 42 suns, which is the illumination concentration used for the 48 h electrolysis. The key performance parameters are included in the figure. I–V curves were collected using both forward and backward voltage sweeps. These measurements generated identical results; thus, this figure shows only the forward sweep results.

35% 250 180 30% 160

200 25% 140 120 20% 32% 180 PV BOO 100 150 170 PV EOO 15% 30% 2 Electrolysers BOO 80 160 Current (mA) 2 Electrolysers EOO 28% 60 Current (mA) 10% 100 0 10 20 30 40 50 40 Solar-to-hydrogen efficiency 5% 20 50 0% 0 0 5 10 15 20 25 30 35 40 45 50 0 Time (h) 2.4 2.6 2.8 3.0 3.2 Figure 4 | STH efficiency. The STH efficiency of the PV-electrolysis system Voltage (V) was measured over a 48 h continuous operation. The right vertical axis Figure 3 | PV cell and PEM electrolyser performance at the beginning shows the current passing through the dual electrolyser and the left vertical and end of operation. The I–V characteristics of the triple-junction solar cell axis shows the corresponding STH efficiency. The inset highlights a smaller and the dual-electrolyser at both beginning-of-operation (BOO) and end-of- y axis range for improved clarity. operation (EOO). The blue and cyan curves are the solar cell I–V curves under 42 suns at BOO and EOO, respectively. The dark and light red curves mance is most likely to be due to catalyst degradation (loss of are the I–V curves of the dual-electrolysers at BOO and EOO, respectively. catalyst material) as well as potential membrane degradation or The BOO electrolyser I–V was measured as a single cyclic voltammogram to contaminants introduced from recirculating water for extended minimize catalyst degradation before system operation. The EOO periods of time. In addition, small amounts of water were electrolyser I–V is presented as the average current for 2 min holds at each observed in the cathode outlet stream, suggesting water crossover potential, shown with error bars indicating 1 s.d. and a connecting line to and buildup on the cathode side, which could have blocked guide the eye. catalyst active sites and decreased over time. Although the increase of VOP over VMPP resulted in a slight decrease of IOP and the STH efficiency over the 48 h test cells, we used the ratio of the short circuit currents to calculate operation, the solar cell and dual-electrolyser were well matched that the cell was illuminated with 42 suns white light under the near the maximum power point of the solar cell, ensuring concentration condition (calculation details provided in PV-electrolysis performance near the optimum. Supplementary Note 1)32,33. At the maximum power point, the cell voltage (VMPP) output is 2.91 V and the current density (JMPP) is 565.9 mA cm 2. The measured solar-to-electricity efficiency is STH efficiency. Figure 4 shows the electrolysis current and the 39% under these operating conditions. The I–V characteristics of corresponding STH efficiency through the 48 h experiment. The the dual electrolyser and the solar cell, measured before and after operating current decreased by only 10% over this period from an the 48 h operation, are shown in Fig. 3. The cross-point of a dual- initial value of 177 mA to a final value of 160 mA. The STH electrolyser I–V curve and a solar cell I–V curve is the system- efficiency was calculated by multiplying two times the thermo- coupling point and indicates the operating voltage (VOP) and dynamic potential (Vredox), the electrolysis current (IWE) and the current (IOP) for the system. It can be seen that at the beginning Faradaic efficiency for hydrogen evolution (ZF), then dividing by 34 of operation, VOP is slightly lower than VMPP. At the end of the input light power (Pin) . The factor of two comes from the operation, as the activity of the dual-electrolyser decreases, VOP is fact that there are two electrolysers connected in series; thus, the slightly higher than VMPP. The decrease in electrolyser perfor- electrolysis current is driving two electrolysis reactions at the

NATURE COMMUNICATIONS | 7:13237 | DOI: 10.1038/ncomms13237 | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13237 same time. a result, the STH efficiency numbers we have reported for this device represent a limiting upper bound in regard to any 2VredoxIWEZ STH ¼ F ð1Þ additional auxiliary energy inputs that would be required for any P in commercially scaled water-splitting system. In this system, the thermodynamic potential for the water- We note that the laboratory-scale device described in this study splitting reaction is 1.18 V, as the dual electrolyser was operating is constructed from components that could be scaled up to larger at 80 °C. A Pin value of 1,328 mW was obtained by multiplying PV-electrolysis installations. Optimizing the PV-electrolysis the operational concentration (42 suns) by the 1 sun illumination system design will be crucial for enabling the deployment of intensity (100 cm 2) and the solar cell area (0.316 cm2). economical large-scale installations. Appropriate designs will The Faradaic efficiency of the electrolyser was measured at minimize the heating, cooling and pumping energy inputs. 20 min intervals several times throughout the 48 h experiment Most commercial concentrated PV modules use passive cooling (photographs of the measurement apparatus are shown in methods to regulate the PV cell temperature35 and PEM electro- Supplementary Fig. 2). The results of the Faradaic efficiency lysers will naturally run at temperatures well above ambient measurements and calculations indicate that the electrolyser is conditions, as the excess energy from overpotentials at the anode selective for producing H2 and O2 (results shown in and cathode is dispersed as heat. If needed, additional heat could Supplementary Fig. 3). We assume that our system has nearly be supplied by the sunlight that is not converted to electricity. It 100% Faradaic efficiency for hydrogen production, consistent might also be desirable to operate the PV at a temperature with other studies of PEM electrolysers23–25. Further details about 425 °C. Although this would decrease the PV’s open circuit the product quantification measurements are in Supplementary voltage, it would increase the current output and decrease the Note 2 and procedure for STH calculations are provided in energy input needed to cool the PV (if active cooling methods Supplementary Note 3. were employed). This could result in an increased STH efficiency Based on equation (1), the system reached an average STH if the PV still produced enough voltage to drive the dual- efficiency of over 31% in the first 20 min of operation and electrolyser. The PV operating temperature and other operation maintained an STH efficiency of 430% for the first 20þ h. The parameters pose tradeoffs to be optimized. Overall, the system STH efficiency slowly decreased over 48þ h of testing and design and operating conditions must be chosen to minimize the averaged B28% for the last 20 min of operation. The average levelized cost of hydrogen produced over the operating lifetime of STH efficiency achieved over the entire 48þ h of operation was the system. Although the components used in this device are 30%. The decrease in STH over 48 h of operation is mainly due to expensive, including the III–V PV fabrication and the precious an increase of VOP over the maximum power point primarily metal catalysts, the very high STH efficiency achieved shows that caused by a decrease in performance of the electrolysers over this strategy may have the potential to produce hydrogen at low time, an issue that is readily addressable through electrolyser cost, in particular if lower-cost materials and fabrication methods development, as commercial electrolysers have demonstrated tens can be developed. Technoeconomic models of large scale, of thousands of hours of stable operation. Although electrolyser centralized solar H2 production facilities suggest that achieving performance accounts for most of the decrease in PV-electrolysis high STH efficiency is one of the most important factors in efficiency over time, Fig. 3 also shows that the current of the solar reducing the cost of H2, potentially even more important than cell was slightly lower at the end of operation compared with the reducing the cost of the absorber or catalyst materials13,14. beginning of operation, which is due to a slight decrease of the Continued development of lower cost, higher efficiency PV cells incident light intensity. This contributed approximately one and electrolysers, optimized PV-electrolysis system designs and percentage point to the total STH loss, as the STH calculation was technoeconomic models to predict hydrogen costs are all performed using the initial incident light intensity; thus, the important subjects for continuing research. overall STH efficiencies reported are conservative values. In summary, we report a PV-electrolysis system that demon- strated an average STH efficiency of 30% over a 48 h period of Discussion continuous operation. This is the highest STH efficiency reported The efficiency analysis assumes that the light incident on the PV to date and the first solar water splitting system that demonstrates cell is the only energy input into the system. In reality, there can a STH efficiency reaching 30% or higher. Coupling a high- be additional energy inputs, including the energy required to cool efficiency multi-junction solar cell with two electrolysers in series the PV, heat the PEM electrolysers and pump water into the is an effective way to minimize the excessive voltage generated by electrolysers. In a concentrated PV device illuminated by natural a multi-junction solar cell, allowing for greater utilization of the sunlight, there may also be optical losses in the lenses or mirrors high-efficiency PV for water splitting. This system also primarily used to concentrate the incident sunlight onto the PV cell, which uses commercially available components, suggesting that could reduce the system efficiency. We have chosen to neglect similar techniques could be implemented on a large scale. The these factors in this study for several reasons, among them to demonstration of this high-efficiency system is an important step allow for direct comparisons with results from previous studies. closer to the US Department of Energy technology and cost goals, We note in general that laboratory-scale systems are designed and shows great opportunities for solar energy storage and H2 differently than commercial-scale systems and the losses are not production with solar water splitting. identical. For instance, in this study, lenses or mirrors were not needed to achieve the simulated concentrated solar light; thus, there are no optical losses to measure. In addition, commercial- Methods scale PV systems and PEM electrolysers have been designed to Solar simulator calibration and Solar Cell characterization. For the 1 sun measurement, we used a solar simulator (ABET Technologies, Model Sun2000) minimize energy losses involving cooling, heating, pumping and equipped with a 550 W xenon lamp as a light source. The GaInP/GaAs/GaInNAsSb so on; thus, the losses in a laboratory-scale demonstration system multi-junction solar cell (0.316 cm2 area) was manufactured by Solar Junction. are less relevant, in addition to being more difficult to accurately As the cells were operated under concentrated sunlight, the calibration of the quantify. Ultimately, the methods used to supply the aforemen- solar simulator for 1 sun conditions (100 mW cm 2) was carried out using the AM 1.5 Direct spectrum (ASTM G173). Although reference solar cells can be used to tioned additional energy inputs in a laboratory-scale system are adjust a simulator for appropriate total power output, spectral control is crucial for not optimized; thus, they are not an accurate representation of accurate multi-junction cell measurements. For this reason, the external quantum what could be achieved in large-scale PV-electrolysis systems. As efficiency (EQE) of each sub-junction of the multi-junction stack was measured

4 NATURE COMMUNICATIONS | 7:13237 | DOI: 10.1038/ncomms13237 | www.nature.com/naturecommunications NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13237 ARTICLE using a grating monochromator (Newport CS260) calibrated with silicon and cell and the solar simulator was adjusted so that B42 suns of solar concentration germanium photodetectors (Newport 918D-UV, 918D-IR). All light sources and was achieved. At this concentration, the solar cell output a short circuit photodetectors were calibrated by the manufacturers before the experiment. These photocurrent of 184 mA (B583 mA cm 2) and aligned the solar cell I–V curve for EQE measurements were integrated with the AM 1.5D spectrum to determine an optimal operation point to match the electrode size and electrolyser capacity. short-circuit current densities and to understand the current-limiting junction of To begin operation, the shutter of the solar simulator was opened. The system the cells. The EQE of each sub-junction and the AM 1.5D spectrum are shown current was recorded continuously by the potentiostat and these data were used to overlaid in Supplementary Fig. 1. The cells were all top-junction limited, which calculate the STH efficiency as a function of time as shown in Fig. 4. The system allowed the simulator to be tuned without luminescent coupling impacting the was run continuously for 48 h without interruption or modification. Periodically current27.The1sunI–V characteristics of the cell were measured using this throughout the experiment, the gas products from the cathodes and anodes of the condition. These data are shown in Fig. 2a. electrolysers were collected using volume displacement devices to calculate the For the PV-electrolysis measurement, we used a multi-sun solar simulator Faradaic efficiency. At the end of the 48 h operation, the shutter of the solar (Newport, Model 66921) with a 1,000 W xenon lamp as the white light source. simulator was closed. A water filter was applied to the light beam, to remove the excessive infrared component from the lamp spectrum and to better match the AM 1.5D spectrum. The cell package was mounted onto a water-cooling stage such that a surface temperature of 25 °C was maintained. The distance between the cell and the lamp Data availability. The data that support the findings of this study are available was adjusted to achieve the desired short circuit current (JSC). The intensity of the from the authors on request. concentrated white light was determined from the ratio of the JSC under concentration to the JSC under 1 sun illumination, consistent with standard practices for characterizing concentrated PV cells32,33. No concentrator optics were placed between the simulator and the cell and therefore the possible effects of References concentrator optics were not considered in efficiency calculation. As the cell was 1. Metz, A. et al. International Technology Roadmap for Photovoltaic (ITRPV): cut and installed in a standard CPV cell package, the whole cell area is active under 2014 Results, Revision 1. http://itrpv.net/ (2015). illumination; therefore, no aperture or mask was used. Numerical modelling based 2. Lewis, N. S. Toward cost-effective solar energy use. Science 315, 798–801 on the junction ideality factor was later conducted to determine whether open- (2007). circuit voltage (VOC) inflation due to the mismatch between the solar simulator 3. Abbasi, T. & Abbasi, S. ‘Renewable’ hydrogen: prospects and challenges. Renew. spectrum and the AM 1.5D spectrum resulted in STH inflation36–38. The results of Sustainable Energy Rev. 15, 3034–3040 (2011). these calculations show that the lamp spectrum mismatch did not cause significant 4. Balat, M. Potential importance of hydrogen as a future solution to STH inflation in our experiments (details provided in Supplementary Note 4 and environmental and transportation problems. Int. J. Hydrogen Energy 33, Supplementary Table 1). 4013–4029 (2008). During the duration of the experiment the solar cell performance was very 5. Nowotny, J., Sorrell, C. C., Sheppard, L. R. & Bak, T. Solar-hydrogen: stable, as expected for III–V solar cells. No hysteresis or time-transient behaviour environmentally safe fuel for the future. Int. J. Hydrogen Energy 30, 521–544 was observed during the I–V measurement. The I–V characteristics results in Fig. 2 (2005). 1 were measured with a forward voltage sweep rate of 50 mV s and a sampling 6. Turner, J. et al. Renewable hydrogen production. Int. J. Energy Res. 32, 379–407 period of 0.02 s. Changing the sweep rate, direction or sampling rate did not (2008). generate a noticeable difference in the results. 7. Walter, M. G. et al. Solar water splitting cells. Chem. Rev. 110, 6446–6473 (2010). 8. Yilmaz, F., Balta, M. T. & Selbas¸, R. A review of solar based hydrogen Electrolyser fabrication and characterization. Membrane electrode assemblies production methods. Renew. Sustainable Energy Rev. 56, 171–178 were fabricated using a conventional catalyst-coated membrane technique. Nafion : 115 membranes purchased from FuelCellsEtc were cut into 3.5 cm 3.5 cm2 pieces. ð2016Þ These membranes were pretreated by soaking in 3% H O at 80 °C for 1 h, then 9. Ruth, M. & Joseck, F. Hydrogen Threshold Cost Calculation (United States 2 2 Department of Energy, 2011). soaking in 0.5 M H2SO4 at 80 °C for 1 h and finally soaking in Millipore (18.2 MO cm) water at 80 °C for 1 h. The membranes were removed from the water 10. Miller, E., Ainscough, C. & Talapatra, A. Hydrogen Production Status 2006-2013 and blotted dry before the catalyst was deposited. Next, a Pt black catalyst (ETEK) (United States Department of Energy, 2014). and Nafion 117 ionomer solution (Aldrich) were mixed in a 3:1 weight ratio. 11. Holladay, J. D., Hu, J., King, D. L. & Wang, Y. An overview of hydrogen Separately, an Ir black catalyst (Premetek) and Nafion 117 ionomer solution production technologies. Catal. Today 139, 244–260 (2009). (Aldrich) were also mixed in a 3:1 weight ratio. The Pt and Ir catalyst/ionomer 12. Bonke, S. A., Wiechen, M., MacFarlane, D. R. & Spiccia, L. Renewable fuels mixtures were both dispersed in 4:1 volume ratio mixtures of isopropanol and from : towards practical artificial . water. The catalyst/ionomer solutions were sonicated for several minutes and then Energy Environ. Sci. 8, 2791–2796 (2015). deposited onto opposite sides of the Nafion membranes by spray casting. The Pt 13. Pinaud, B. A. et al. Technical and economic feasibility of centralized facilities catalyst was loaded on the cathode side at 0.5 mg cm 2 and the Ir catalyst was for solar hydrogen production via and photoelectrochemistry. loaded on the anode side at 2 mg cm 2 over a 2.5 cm 2.5 cm area for a total Energy Environ. Sci. 6, 1983–2002 (2013). device active area of 6.25 cm2. This catalyst-coated membrane was pressed between 14. Dumortier, M., Tembhurne, S. & Haussener, S. Holistic design guidelines for paper (Sigracet GDL 35BC, Ion Power) on the cathode side and Ti mesh solar hydrogen production by photo-electrochemical routes. Energy Environ. (Dexmet) on the anode side. Two identical assemblies prepared in this manner Sci. 8, 3614–3628 (2015). were loaded into cell assemblies (5 cm2, Technologies, Inc.), which were 15. Ager, J. W., Shaner, M., Walczak, K., Sharp, I. D. & Ardo, S. Experimental maintained at a temperature of 80 °C for all measurements. The two electrolysers demonstrations of spontaneous, solar-driven photoelectrochemical water were connected in series. Millipore water (18.2 MO cm) preheated to 80 °C was fed splitting. Energy Environ. Sci. 8, 2811–2824 (2015). into the anode side of the first electrolyser; there was no input to the cathode side. 16. Khaselev, O. & Turner, J. A. A monolithic photovoltaic-photoelectrochemical The cathode and anode outputs of the first electrolyser were connected to the device for hydrogen production via water splitting. Science 280, 425–427 cathode and anode inputs of the second electrolyser and both outputs of the second (1998). electrolyser were collected; thus, the H2 and O2 products could be quantified using 17. Bolton, J. R., Strickler, S. J. & Connolly, J. S. Limiting and realizable efficiencies a volume displacement Faradaic efficiency measurement apparatus. of solar photolysis of water. Nature 316, 495–500 (1985). 18. Hu, S., Xiang, C., Haussener, S., Berger, A. D. & Lewis, N. S. An analysis of the PV-electrolysis system operation. To construct the PV-electrolysis system, the optimal band gaps of light absorbers in integrated tandem triple junction PV cell, the two electrolysers and a potentiostat (BioLogic, VMP3) photoelectrochemical water-splitting systems. Energy Environ. Sci. 6, were connected in series as follows: the working electrode port of the potentiostat 2984–2993 (2013). was connected to the bottom contact of the solar cell, the top contact of the solar 19. Rocheleau, R. E. & Miller, E. L. Photoelectrochemical production of hydrogen: cell was connected to the anode of the first electrolyser, the cathode of the first engineering loss analysis. Int. J. Hydrogen Energy 22, 771–782 (1997). electrolyser was connected to the anode of the second electrolyser and the cathode 20. Seitz, L. C. et al. Modeling practical performance limits of photoelectrochemical of the second electrolyser was connected to the counter electrode port of the water splitting based on the current state of materials research. ChemSusChem. potentiostat. The potentiostat reference lead was connected to the counter electrode 7, 1372–1385 (2014). lead so that it could measure the current passing through the closed system; no 21. Ayers, K. E. et al. Research advances towards low cost, high efficiency PEM additional potential was applied. All electrical connections were made with stan- electrolysis. ECS Transact. 33, 3–15 (2010). dard copper cables, which introduced negligible resistance compared with other 22. Carmo, M., Fritz, D. L., Merge, J. & Stolten, D. A comprehensive components of the system. review on PEM water electrolysis. Int. J. Hydrogen Energy 38, 4901–4934 Before the start of the operation, preheated Millipore water was purged with H2 (2013). and O2, and pumped into the two electrolysers with a Chem-tech Series 100 pump 23. Fujii, K. et al. Characteristics of hydrogen generation from water splitting by at a flow rate of 42 ml min 1. The solar cell was kept at 25 °C on a water cooler polymer electrolyte electrochemical cell directly connected with concentrated stage and positioned under the multi-sun solar simulator. The distance between the photovoltaic cell. Int. J. Hydrogen Energy 38, 14424–14432 (2013).

NATURE COMMUNICATIONS | 7:13237 | DOI: 10.1038/ncomms13237 | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13237

24. Nakamura, A. et al. A 24.4% solar to hydrogen energy conversion efficiency by Water Splitting Technologies’, which was co-sponsored by the National Science combining concentrator photovoltaic modules and electrochemical cells. Appl. Foundation, Division of Chemical, Bioengineering, Environmental and Transport Phys. Express 8, 107101 (2015). Systems (CBET), and the US Department of Energy, Office of Energy Efficiency and 25. Peharz, G., Dimroth, F. & Wittstadt, U. Solar hydrogen production by water Renewable Energy, Fuel Cell Technologies Office. J.D.B. and J.J. acknowledge support splitting with a conversion efficiency of 18%. Int. J. Hydrogen Energy 32, from Stanford Graduate Fellowships. L.C.S. acknowledges support from the DARE 3248–3252 (2007). Doctoral Fellowship provided by the Vice Provost for Graduate Education at Stanford 26. Licht, S. et al. Efficient solar water splitting, exemplified by RuO2-catalyzed University. J.D.B. and L.C.S. acknowledge support from the National Science Foundation AlGaAs/Si . J. Phys. Chem. B 104, 8920–8924 (2000). Graduate Research Fellowship Program. We thank Solar Junction Corporation for 27. Green, M. A., Emery, K., Hishikawa, Y., Warta, W. & Dunlop, E. D. Solar cell providing the solar cell samples. We also thank Thomas Hellstern, Pongkarn Chak- efficiency tables (version 47). Prog. Photovolt. Res. Appl. 24, 3–11 (2016). thranont, Vijay Parameshwaran and Ariel Jackson for helpful discussions. 28. De Vos, A. Detailed balance limit of the efficiency of tandem solar cells. J. Phys. D Appl. Phys. 13, 839 (1980). Author contributions 29. Li, J. & Wu, N. Semiconductor-based photocatalysts and photoelectrochemical J.J., L.C.S. and J.D.B. contributed equally to this work. J.J., L.C.S., J.D.B., Y.H., J.S.H. and cells for generation: a review. Catal. Sci. Technol. 5, 1360–1384 T.F.J. conceived the study and designed the experiments. J.J., L.C.S. and J.D.B. performed (2015). the experiments. T.B. and Y.C. contributed to PV cell preparation and J.W.D.N. 30. Bak, T., Nowotny, J., Rekas, M. & Sorrell, C. C. Photo-electrochemical contributed to electrolyser preparation. All authors contributed to analysing the data hydrogen generation from water using solar energy. Materials-related aspects. and writing the manuscript. Int. J. Hydrogen Energy 27, 991–1022 (2002). 31. Sabnis, V., Yuen, H. & Wiemer, M. High-efficiency multijunction solar cells employing dilute nitrides. AIP Conf. Proc. 1477, 14–19 (2012). 32. Osterwald, C. Translation of device performance measurements to reference Additional information conditions. Solar Cells 18, 269–279 (1986). Supplementary Information accompanies this paper at http://www.nature.com/ 33. Osterwald, C., Wanlass, M., Moriarty, T., Steiner, M. & Emergy, K. in IEEE 40th naturecommunications Photovoltaic Specialist Conference (PVSC) 2616 (Denver, 2014). Competing financial interests: The authors declare no competing financial 34. Chen, Z. et al. Accelerating materials development for photoelectrochemical interests. hydrogen production: standards for methods, definitions, and reporting protocols. J. Mater. Res. 25, 3–16 (2010). Reprints and permission information is available online at http://npg.nature.com/ 35. Du, D., Darkwa, J. & Kokogiannakis, G. Thermal management systems for photo- reprintsandpermissions/ voltaics (PV) installations: a critical review. Solar Energy 97, 238–254 (2013). 36. Geisz, J. F. et al. Generalized optoelectronic model of series-connected How to cite this article: Jia, J. et al. Solar water splitting by photovoltaic-electrolysis multijunction solar cells. IEEE J. Photovolt. 5, 1827–1839 (2015). with a solar-to-hydrogen efficiency over 30%. Nat. Commun. 7, 13237 37. Meusel, M., Adelhelm, R., Dimroth, F., Bett, A. W. & Warta, W. Spectral doi: 10.1038/ncomms13237 (2016). mismatch correction and spectrometric characterization of monolithic III–V multi-junction solar cells. Prog. Photovolt. Res. Appl. 10, 243–255 (2002). This work is licensed under a Creative Commons Attribution 4.0 38. Siefer, G. et al. Influence of the simulator spectrum on the calibration of multi- International License. The images or other third party material in this junction solar cells under concentration. In Photovoltaic Specialists Conference, article are included in the article’s Creative Commons license, unless indicated otherwise 2002. Conference Record of the Twenty-Ninth IEEE 836–839 (2002). in the credit line; if the material is not included under the Creative Commons license, users will need to obtain permission from the license holder to reproduce the material. Acknowledgements To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ This study presents results from an NSF project (award number CBET-1433442) competitively selected under the solicitation ‘NSF 14-15: NSF/Department of Energy Partnership on Advanced Frontiers in Renewable Hydrogen Fuel Production via Solar r The Author(s) 2016

6 NATURE COMMUNICATIONS | 7:13237 | DOI: 10.1038/ncomms13237 | www.nature.com/naturecommunications